Bialystok Outline Annotated Cervical Cytology Dataset

The Białystok dataset contains 162 images measuring 3500×3500 px, with 2,419 cells annotated in accordance with the Bethesda system. The images are derived from whole-slide images (WSI) of routine cervical smears.
The dataset is described in detail in the article „Cytoplasm and nuclei as a basis for Bethesda cell cluster classification in cervical smears”.
To access the dataset, please fill out the form.
The dataset presents a realistic variety of smears, ranging from simple and easily distinguishable cells to challenging examples filled with stained mucus, large dark cell clusters, and images abundant with neutrophils obscuring the view.

The dataset contains artefacts typically found in cytology smears:

The dataset is annotated with cytoplasm maps divided into the 6 Bethesda categories and two unidentifiable categories: Unidentified Cells and Unidentified Cell Clusters. The “unidentified” category is a result of the specific nature of the cytotechnologist’s work. The spectrum of cell condition is continuous. Therefore, under certain conditions, the cytotechnologist cannot make a definitive decision. Such cells are still marked in the dataset to enable the segmentation of all cells in the image.

In conclusion, the dataset provides:
– Realistic WSI fragments of routine smears
– Annotations consistent with a standard reporting system
– Suitable for segmentation, classification, and detection
– An advanced benchmark